jamilbk

13 ms·

1. 121.▲
   by jamilbk 9y ago

   Another important factor to consider is Vega supports double speed FP16 operations[1] and some ML frameworks are already beginning to optimize for that[2], so that's almost 24 TFLOPS of training compute for ~ $400 USD on the RX Vega 56
2. 122.▲
   by jamilbk 9y ago

   The problem with “Don’t be evil” is that evil is subjective.
3. 123.▲
   Open Source Deep Learning on AMD and Beyond(vertex.ai)

   2 points by jamilbk 9y ago|0 comments